Performs Gauss-Newton weighted least squares computation. More...

#include <wls.h>

Public Member Functions
	WLS (int size=0)
	Default constructor or construct with the number of dimensions for the Dynamic case.

void	clear ()
	Clear all the measurements and apply a constant regularisation term.

void	add_prior (Precision val)
	Applies a constant regularisation term. More...

template<class B2 >
void	add_prior (const Vector< Size, Precision, B2 > &v)
	Applies a regularisation term with a different strength for each parameter value. More...

template<class B2 >
void	add_prior (const Matrix< Size, Size, Precision, B2 > &m)
	Applies a whole-matrix regularisation term. More...

template<class B2 , class P2 >
void	add_mJ (Precision m, const Vector< Size, P2, B2 > &J, Precision weight=1)
	Add a single measurement. More...

template<class VB , int S2, class B2 >
void	add_mJ (const Vector< S2, Precision, VB > &m, const Matrix< S2, Size, Precision, B2 > &J, Precision weight=1)
	Add a several measurements measurement. More...

template<int N, class B1 , class B2 , class B3 >
void	add_mJ (const Vector< N, Precision, B1 > &m, const Matrix< Size, N, Precision, B2 > &J, const Matrix< N, N, Precision, B3 > &invcov)
	Add multiple measurements at once (much more efficiently) More...

template<int N, class B1 , class B2 , class B3 >
void	add_mJ_rows (const Vector< N, Precision, B1 > &m, const Matrix< N, Size, Precision, B2 > &J, const Matrix< N, N, Precision, B3 > &invcov)
	Add multiple measurements at once (much more efficiently) More...

template<int N, typename B1 >
void	add_sparse_mJ (const Precision m, const Vector< N, Precision, B1 > &J1, const int index1, const Precision weight=1)
	Add a single measurement at once with a sparse Jacobian (much, much more efficiently) More...

template<int N, int S1, class P1 , class P2 , class P3 , class B1 , class B2 , class B3 >
void	add_sparse_mJ_rows (const Vector< N, P1, B1 > &m, const Matrix< N, S1, P2, B2 > &J1, const int index1, const Matrix< N, N, P3, B3 > &invcov)
	Add multiple measurements at once with a sparse Jacobian (much, much more efficiently) More...

template<int N, int S1, int S2, class B1 , class B2 , class B3 , class B4 >
void	add_sparse_mJ_rows (const Vector< N, Precision, B1 > &m, const Matrix< N, S1, Precision, B2 > &J1, const int index1, const Matrix< N, S2, Precision, B3 > &J2, const int index2, const Matrix< N, N, Precision, B4 > &invcov)
	Add multiple measurements at once with a sparse Jacobian (much, much more efficiently) More...

void	compute ()
	Process all the measurements and compute the weighted least squares set of parameter values stores the result internally which can then be accessed by calling get_mu()

void	operator+= (const WLS &meas)
	Combine measurements from two WLS systems. More...

Matrix< Size, Size, Precision > &	get_C_inv ()
	Returns the inverse covariance matrix.

const Matrix< Size, Size, Precision > &	get_C_inv () const
	Returns the inverse covariance matrix.

Vector< Size, Precision > &	get_mu ()
	Returns the update. With no prior, this is the result of \(J^\dagger e\).

const Vector< Size, Precision > &	get_mu () const
	Returns the update. With no prior, this is the result of \(J^\dagger e\).

Vector< Size, Precision > &	get_vector ()
	Returns the vector \(J^{\mathsf T} e\).

const Vector< Size, Precision > &	get_vector () const
	Returns the vector \(J^{\mathsf T} e\).

Decomposition< Size, Precision > &	get_decomposition ()
	Return the decomposition object used to compute \((J^{\mathsf T} J + P)^{-1}\).

const Decomposition< Size, Precision > &	get_decomposition () const
	Return the decomposition object used to compute \((J^{\mathsf T} J + P)^{-1}\).

Detailed Description

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>
class TooN::WLS< Size, Precision, Decomposition >

Performs Gauss-Newton weighted least squares computation.

Parameters

Size	The number of dimensions in the system
Precision	The numerical precision used (double, float etc)
Decomposition	The class used to invert the inverse Covariance matrix (must have one integer size and one typename precision template arguments) this is Cholesky by default, but could also be SQSVD

Member Function Documentation

◆ add_mJ() [1/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<class B2 , class P2 >

void TooN::WLS< Size, Precision, Decomposition >::add_mJ	(	Precision	m,
		const Vector< Size, P2, B2 > &	J,
		Precision	weight = `1`
	)

inline

Add a single measurement.

Parameters

m	The value of the measurement
J	The Jacobian for the measurement \(\frac{\partial\text{m}}{\partial\text{param}_i}\)
weight	The inverse variance of the measurement (default = 1)

◆ add_mJ() [2/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<class VB , int S2, class B2 >

void TooN::WLS< Size, Precision, Decomposition >::add_mJ	(	const Vector< S2, Precision, VB > &	m,
		const Matrix< S2, Size, Precision, B2 > &	J,
		Precision	weight = `1`
	)

inline

Add a several measurements measurement.

Parameters

m	The value of the measurement
J	The Jacobian for the measurement \(\frac{\partial\text{m}}{\partial\text{param}_i}\)
weight	The inverse variance of the measurement (default = 1)

◆ add_mJ() [3/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<int N, class B1 , class B2 , class B3 >

void TooN::WLS< Size, Precision, Decomposition >::add_mJ	(	const Vector< N, Precision, B1 > &	m,
		const Matrix< Size, N, Precision, B2 > &	J,
		const Matrix< N, N, Precision, B3 > &	invcov
	)

inline

Add multiple measurements at once (much more efficiently)

Parameters

m	The measurements to add
J	The Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
invcov	The inverse covariance of the measurement values

◆ add_mJ_rows()

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<int N, class B1 , class B2 , class B3 >

void TooN::WLS< Size, Precision, Decomposition >::add_mJ_rows	(	const Vector< N, Precision, B1 > &	m,
		const Matrix< N, Size, Precision, B2 > &	J,
		const Matrix< N, N, Precision, B3 > &	invcov
	)

inline

Add multiple measurements at once (much more efficiently)

Parameters

m	The measurements to add
J	The Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
invcov	The inverse covariance of the measurement values

◆ add_prior() [1/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

void TooN::WLS< Size, Precision, Decomposition >::add_prior ( Precision val )

inline

Applies a constant regularisation term.

Equates to a prior that says all the parameters are zero with \(\sigma^2 = \frac{1}{\text{val}}\).

Parameters

val	The strength of the prior

◆ add_prior() [2/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<class B2 >

void TooN::WLS< Size, Precision, Decomposition >::add_prior ( const Vector< Size, Precision, B2 > & v )

inline

Applies a regularisation term with a different strength for each parameter value.

Equates to a prior that says all the parameters are zero with \(\sigma_i^2 = \frac{1}{\text{v}_i}\).

Parameters

v	The vector of priors

◆ add_prior() [3/3]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<class B2 >

void TooN::WLS< Size, Precision, Decomposition >::add_prior ( const Matrix< Size, Size, Precision, B2 > & m )

inline

Applies a whole-matrix regularisation term.

This is the same as adding the \(m\) to the inverse covariance matrix.

Parameters

m	The inverse covariance matrix to add

◆ add_sparse_mJ()

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<int N, typename B1 >

void TooN::WLS< Size, Precision, Decomposition >::add_sparse_mJ	(	const Precision	m,
		const Vector< N, Precision, B1 > &	J1,
		const int	index1,
		const Precision	weight = `1`
	)

inline

Add a single measurement at once with a sparse Jacobian (much, much more efficiently)

Parameters

m	The measurements to add
J1	The first block of the Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
index1	starting index for the first block
invcov	The inverse covariance of the measurement values

◆ add_sparse_mJ_rows() [1/2]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<int N, int S1, class P1 , class P2 , class P3 , class B1 , class B2 , class B3 >

void TooN::WLS< Size, Precision, Decomposition >::add_sparse_mJ_rows	(	const Vector< N, P1, B1 > &	m,
		const Matrix< N, S1, P2, B2 > &	J1,
		const int	index1,
		const Matrix< N, N, P3, B3 > &	invcov
	)

inline

Add multiple measurements at once with a sparse Jacobian (much, much more efficiently)

Parameters

m	The measurements to add
J1	The first block of the Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
index1	starting index for the first block
invcov	The inverse covariance of the measurement values

◆ add_sparse_mJ_rows() [2/2]

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

template<int N, int S1, int S2, class B1 , class B2 , class B3 , class B4 >

void TooN::WLS< Size, Precision, Decomposition >::add_sparse_mJ_rows	(	const Vector< N, Precision, B1 > &	m,
		const Matrix< N, S1, Precision, B2 > &	J1,
		const int	index1,
		const Matrix< N, S2, Precision, B3 > &	J2,
		const int	index2,
		const Matrix< N, N, Precision, B4 > &	invcov
	)

inline

Add multiple measurements at once with a sparse Jacobian (much, much more efficiently)

Parameters

m	The measurements to add
J1	The first block of the Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
index1	starting index for the first block
J2	The second block of the Jacobian matrix \(\frac{\partial\text{m}_i}{\partial\text{param}_j}\)
index2	starting index for the second block
invcov	The inverse covariance of the measurement values

◆ operator+=()

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>

void TooN::WLS< Size, Precision, Decomposition >::operator+= ( const WLS< Size, Precision, Decomposition > & meas )

inline

Combine measurements from two WLS systems.

Parameters

meas	The measurements to combine with

The documentation for this class was generated from the following file:

TooN/wls.h

Public Member Functions

Detailed Description

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky> class TooN::WLS< Size, Precision, Decomposition >

Member Function Documentation

◆ add_mJ() [1/3]

◆ add_mJ() [2/3]

◆ add_mJ() [3/3]

◆ add_mJ_rows()

◆ add_prior() [1/3]

◆ add_prior() [2/3]

◆ add_prior() [3/3]

◆ add_sparse_mJ()

◆ add_sparse_mJ_rows() [1/2]

◆ add_sparse_mJ_rows() [2/2]

◆ operator+=()

template<int Size = Dynamic, class Precision = DefaultPrecision, template< int DecompSize, class DecompPrecision > class Decomposition = Cholesky>
class TooN::WLS< Size, Precision, Decomposition >